From Science to Capital: How the Best Founders Actually Fund Transformative Innovation

July 18, 2026 | 8 min read

The most common story told about how innovation gets funded is a fiction. A brilliant scientist has an idea, pitches a venture capitalist, raises money, and changes the world.

The actual path is messier, longer, and far more instructive. The people who have actually built transformative companies at the intersection of science and capital share a very different story — one where the real competitive advantage is not the idea but the depth of operational knowledge, the patience to be early, and the willingness to stay when being early is uncomfortable.

Tom Cahill was still a medical student at Duke when he started investing and building companies. He was planning to become an assistant professor when someone told him he would be better off doing something more interesting. That conversation led to New Path Ventures, which backed early breakthroughs in GLP-1 medicine, genomics, and COVID therapeutics.

“Science moves like a punctuated equilibrium — it stays steady and then all of a sudden it changes. The GLP-1s are an example of that. You have to be ahead and see where the system is heading.”

“We knew about GLP-1s in 2018. We knew about sequencing in 2014. A lot of people see things and they don’t make the right bet. If you’re going to be a venture capitalist and you have conviction — you’re gambling on yourself.”

Justin Sherlock did not decide to found Caspian because he researched the market. He founded it because he spent five years at Flexport watching tariffs erode company margins, and noticing that almost none of them were claiming the duty refunds they were legally owed. He became a licensed customs broker. He learned the mechanics from the inside. And when he raised a seed round — before he had a product, a team, or a customer — investors could see he understood the problem at a level no amount of research could replicate:

“Don’t worry about pigeon-holing yourself. Go dive into a space and learn, and then you’ll develop the idea very organically — versus trying to research a market and talk to hundreds of people.”

“We raised our seed round before we had any product or team. I wanted to build the right product without the pressure to commercialize it immediately.”

Mateesh Agarwal came to the United States at 17 to study chemical engineering. His first startup failed at scale. He moved into cloud computing, built Lambda into a company generating nearly half a billion dollars in annual revenue, and then founded Positron AI to build custom silicon for AI inference. The throughline is not a consistent industry — it is a consistent method: find the hardest technical problem in a domain you understand deeply, and build the most efficient solution:

“The why is always about curiosity and wanting to learn more. You’re seeing a technology, you want to be involved in it, and you want to contribute to make it better.”

“We are the only ASIC company that has made it such that you don’t have to make any changes in any line of code to run your same application that you’re running on Nvidia to run on Positron. That’s how you get customers talking to you.”

Tom Cahill, Justin Sherlock, and Mateesh Agarwal came from completely different backgrounds and built completely different companies. But the pattern of how they got from insight to funded company is remarkably consistent.

They spent years inside the problem before they tried to solve it. Cahill was an MD-PhD treating patients before he invested in medicine. Sherlock was a customs broker before he built customs software. Agarwal was building AI clouds before he built AI chips.

They were early — and they stayed early. All three committed to their thesis before the consensus formed, at a time when being right felt indistinguishable from being wrong, and they had the conviction to stay.

They built for the person, not the market. Cahill asks what drug he would take if he had the disease. Sherlock built Caspian around the specific pain he watched hundreds of founders experience. Agarwal builds chips that slot into existing Nvidia workflows without requiring code changes — because he knows his customers will not adopt a solution that requires them to rebuild everything.